移动P2P环境下信任模型与激励机制的研究

发布时间：2018-01-11 17:34

本文关键词：移动P2P环境下信任模型与激励机制的研究　出处：《南京邮电大学》2015年硕士论文　论文类型：学位论文

【摘要】：由于移动P2P(Peer to Peer)网络中节点固有的属性,以及移动P2P网络的拓扑结构随着节点移动而动态变化的特性,使得移动P2P网络面临着严重的安全问题,例如传播恶意虚假文件,滥用网络资源等。同时由于移动P2P网络中节点的自私性,节点不合作的情况较为普遍。构建合理的信任模型能够有效的保障节点之间的交易安全,建立激励机制则能有效的解决P2P网络中的节点的自私行为。本文首先提出基于社会信任补充的信任模型,该模型针对移动P2P网络的中信任信息易缺失的情形,使用社会信任作为补充,社会信任的主要优势在于不需要节点之间有过交易便可以从社会关系的角度对节点是否可信做出评价。而当信任信息可以获得时,则利用传统的综合信任值计算方法:α*直接信任值+β*推荐信任值。在计算推荐信任值时,本文综合节点之间的交易次数和交易满意次数得到推荐信任值,而不是直接综合推荐信息,以免出现由于交易次数过少导致的评价不稳定情况;同时采用三维向量描述节点之间关系,通过计算余弦相似度筛选出可信推荐节点。本文中基于模糊Q学习的激励机制针对节点的不同身份,将不同的节点相关参数作为输入状态变量;依据状态和动作模糊集构建初始模糊规则库;设计即时收益回报函数;描述了基于模糊Q学习的进化博弈具体过程;最后从不同的角度进行仿真,证明了此机制具有较快的Q值收敛速度,并且能够有效的激励节点积极参与到网络活动中。
[Abstract]:The mobile P2P (Peer to Peer) attribute node in the network of natural, as well as the topology of mobile P2P networks with mobile nodes and the dynamic change characteristics of the mobile P2P network is facing serious security problems, such as the spread of malicious false documents, the abuse of cyber source. At the same time due to the mobility of nodes in P2P network selfishness without the cooperation of nodes is more common. The construction of trust model is reasonable and can effectively guarantee the node between the transaction security, the establishment of incentive mechanism can effectively solve the P2P node in the network. This paper proposes the selfish behavior of social trust model based on the model for mobile P2P network trust information missing the use of social trust, social trust as a supplement, the main advantage is that you do not need to have the transaction between nodes can be from the perspective of social relations of the festival is No credible make evaluation. And when the trust information can be obtained when using the traditional comprehensive trust value calculation method: alpha + beta * * direct trust value recommendation trust value. In the calculation of recommended trust value, the comprehensive satisfaction between node number of transactions and transaction times are recommended trust value, rather than directly integrated information recommendation in order to avoid the evaluation of the number of transactions, leading to too little instability; at the same time using the three-dimensional vector to describe the relationship between nodes, by calculating the cosine similarity of selected credible recommender. The incentive mechanism of fuzzy Q learning based on node identity, the nodes associated with different parameters as input variables; on the basis of state and action fuzzy set fuzzy rules of initial design; instant return function; describes the specific process of evolutionary game based on fuzzy Q learning; finally, from different angles The degree of simulation proves that this mechanism has a fast convergence rate of Q value and can effectively encourage nodes to actively participate in the network activity.

【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP393.02

【相似文献】